Learning-Based and Data-Driven TCP Design for Memory-Constrained IoT

@article{Li2016LearningBasedAD,
  title={Learning-Based and Data-Driven TCP Design for Memory-Constrained IoT},
  author={Wei Li and Fan Zhou and Waleed Meleis and Kaushik R. Chowdhury},
  journal={2016 International Conference on Distributed Computing in Sensor Systems (DCOSS)},
  year={2016},
  pages={199-205}
}
  • Wei Li, Fan Zhou, K. Chowdhury
  • Published 26 May 2016
  • Computer Science
  • 2016 International Conference on Distributed Computing in Sensor Systems (DCOSS)
Advances in wireless technology have resulted in pervasive deployment of devices of a high variability in form factors, memory and computational ability. The need for maintaining continuous connections that deliver data with high reliability necessitate re-thinking of conventional design of the transport layer protocol. This paper investigates the use of Q-learning in TCP cwnd adaptation during the congestion avoidance state, wherein the classical alternation of the window is replaced, thereby… 
View on IEEE
krc.coe.neu.edu
23 Citations
Highly Influential Citations
4
Background Citations
17
Methods Citations
9

Figures and Tables from this paper

23 Citations

Multihop Data Delivery Virtualization for Green Decentralized IoT
TLDR
A multihop data delivery virtualization approach which uses multiple one-hop reliable transmissions to perform multihOP data transmissions and can provide a significantly higher throughput and shorter transmission time as compared to the end-to-end approach.
Which AQM fits IoT better?
TLDR
This paper provides extensive simulations within ns-3 to highlight what is the better AQM solution for IoT networks by considering the TCP algorithm and other environmental conditions like the number of active nodes, the network RTT and the presence of multiple RTT flows.
PLink: Efficient Cloud-based Training with Topology-aware Dynamic Hierarchical Aggregation
TLDR
PLink, an optimized communication library that probes the physical network and then generates and executes a fitted hierarchical aggregation plan to take advantage of locality in the datacenter network, delivers up to 3x faster end-to-end training performance for popular DL models on Azure and EC2 compared to the state of the art.
A Survey on Recent Advances in Transport Layer Protocols
TLDR
This paper surveys the main novelties related to transport protocols that have been recently proposed, identifying three main research trends: the evolution of congestion control algorithms, to target optimal performance in challenging scenarios, and the introduction of multipath capabilities at the transport layer.
When Machine Learning Meets Congestion Control: A Survey and Comparison
PLINK: DISCOVERING AND EXPLOITING DATACENTER NETWORK LOCALITY FOR EFFICIENT CLOUD-BASED DISTRIBUTED TRAINING
TLDR
This work proposes PLink, an optimized communication library that probes the physical network then generates and executes a fitted hierarchical aggregation plan to take advantage of topology-induced locality in the datacenter network and evolves the plan to adapt to changing network conditions.
Machine Learning in Congestion Control: A Survey on Selected Algorithms and a New Roadmap to their Implementation
TLDR
This paper surveys selected machine-learning based approaches to CC and proposes a roadmap to their implementation in computer systems, by using dataflow computing and Gallium Arsenide (GaAs) chips.
A Deep Reinforcement Learning Perspective on Internet Congestion Control
TLDR
It is shown that casting congestion control as RL enables training deep network policies that capture intricate patterns in data traffic and network conditions, and leverage this to outperform the state-of-the-art.
Owl: Congestion Control with Partially Invisible Networks via Reinforcement Learning
TLDR
Owl is presented, a transport protocol based on reinforcement learning, whose goal is to select the proper congestion window learning from end-to-end features and network signals, when available, and it is shown that the solution converges to a fair resource allocation after the learning overhead.
A comprehensive survey on machine learning for networking: evolution, applications and research opportunities
TLDR
This survey delineates the limitations, give insights, research challenges and future opportunities to advance ML in networking, and jointly presents the application of diverse ML techniques in various key areas of networking across different network technologies.
...
...

References

SHOWING 1-10 OF 20 REFERENCES
A Machine Learning Approach to End-to-End RTT Estimation and its Application to TCP
TLDR
A novel approach to end-to-end round-trip time (RTT) estimation using a machine-learning technique known as the Experts Framework, which shows a considerable reduction in the number of retransmitted packets and a increase in goodput on more heavily congested scenarios.
PCC: Re-architecting Congestion Control for Consistent High Performance
TLDR
It is proved that PCC converges to a stable and fair equilibrium, and shows consistent and often 10× performance improvement, with better fairness and stability than TCP.
TCP ex machina: computer-generated congestion control
TLDR
A program called Remy is developed that generates congestion-control algorithms to run at the endpoints that outperformed human-designed end-to-end techniques, including TCP Cubic, Compound, and Vegas.
An experimental study of the learnability of congestion control
TLDR
It is found that it may be acceptable to simplify some characteristics of the network---such as its topology---when modeling for design purposes, and some other features, such as the degree of multiplexing and the aggressiveness of contending endpoints, are important to capture in a model.
Sparse Distributed Memories for On-Line Value-Based Reinforcement Learning
TLDR
An algorithm for dynamic on-line allocation and adjustment of memory resources for SDMs is presented, which eliminates the need for choosing the memory size and structure a priori.
A Machine Learning Approach to TCP Throughput Prediction
TLDR
A new lightweight method for TCP throughput prediction using Support Vector Regression (SVR) is described, based on both prior file transfer history and measurements of simple path properties, which shows nearly a threefold improvement over prior history-based methods.
CitySense: An Urban-Scale Wireless Sensor Network and Testbed
TLDR
The goal of CitySense is explicitly not to provide public Internet access, but rather to serve as a new kind of experimental apparatus for urban-scale distributed systems and networking research efforts.
TCP Vegas: End to End Congestion Avoidance on a Global Internet
TLDR
The three key techniques employed by Vegas are described, and the results of a comprehensive experimental performance study, using both simulations and measurements on the Internet, of the Vegas and Reno implementations of TCP are presented.
FAST TCP: Motivation, Architecture, Algorithms, Performance
TLDR
FAST TCP is described, a new TCP congestion control algorithm for high-speed long-latency networks, from design to implementation, and its equilibrium and stability properties are characterized.
RFID Technology for IoT-Based Personal Healthcare in Smart Spaces
TLDR
A survey on the state-of-the-art of RFID for application to body centric systems and for gathering information about the user's living environment is presented.
...
...